agentleFS
adoptaiadoptai/cpa-skills/skills/bank-rec-review/SKILL.md

bank-rec-review

Skill8 starsChanged 4 months agoMITView on GitHub

Audit a bank reconciliation that someone else prepared — independently recompute it, verify every reconciling item against the bank statement and the subsequent-period statement, confirm each item actually cleared, and detect plugs, stale items rolled forward, and reconciling items that exist only on the reconciliation. Use this whenever the user mentions reviewing or testing a bank reconciliation, auditing a client's reconciliation, verifying reconciling items, whether a reconciliation is supported, an outstanding check that never cleared, a suspicious reconciling item, or cash audit procedures. Also trigger on "review the client's bank rec," "test this reconciliation," "did these items clear," "is this reconciliation supported," "audit cash," or when a prepared reconciliation is provided alongside a bank statement. Distinct from preparing a reconciliation — for that, see bank-rec-to-gl. Runs fully local — no client cash data leaves the machine.

  • Installs packages

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    All three survive a review that only checks the arithmetic.

    line 23
  2. Distinctive:

    No clean workpaper unless:

    line 29

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Bank Reconciliation Review

Preparing a reconciliation and reviewing one are different procedures. The preparer starts from two records and works toward agreement. The reviewer starts from a document that already claims agreement, and has to establish whether that claim is true. Those are not the same task, and the review is the easier one to do badly, because a reconciliation that foots looks finished.

A reconciliation can foot perfectly and still be worthless. The three ways:

  1. A fabricated reconciling item. An "outstanding check" or "deposit in transit" that exists only on the reconciliation, inserted to make the two sides agree. It has no counterpart in any bank statement, before or after.
  2. A stale item rolled forward. A genuine item from years ago that has never cleared and never been investigated, carried each month because removing it would break the reconciliation.
  3. A plug with a plausible label. Timing, Bank adjustment, Unidentified deposit, Items in transit. The label is doing the work, not the item.

All three survive a review that only checks the arithmetic. The test that catches all three is subsequent clearance: a real reconciling item appears in the next period's bank statement. A fabricated one never does.

The gate

No clean workpaper unless:

  1. The reconciliation is independently recomputed and the recomputation agrees to what the preparer presented. Do not accept the preparer's subtotals.
  2. The bank balance is agreed to the bank statement and the book balance to the general ledger — as two independent sources. Agreeing the reconciliation to itself proves nothing.
  3. Every reconciling item is supported — traced to the bank statement, the subsequent-period statement, or a document. An item with no support is reported, never accepted.
  4. Subsequent clearance is tested where a subsequent statement is available, and every item that did not clear is reported individually.

Inputs

  1. The reconciliation as prepared, item by item — not just the totals.
  2. The bank statement for the period, with its closing balance.
  3. The general ledger cash balance at the same date, ideally with the account detail.
  4. The subsequent-period bank statement. This is the most important document in the review and it is frequently not requested. Without it, clearance cannot be tested and the review is limited in scope.
  5. Prior-period reconciliations, to detect items rolled forward.
  6. Who prepared and who reviewed it, and when. A reconciliation prepared two months after period end by the person with disbursement authority, unreviewed, is a control finding independent of its arithmetic.

Step 1 — Recompute and test

python3 scripts/rec_review.py \
  --reconciliation client_rec.csv \
  --bank-balance 1284551.09 \
  --gl-balance 1291204.77 \
  --subsequent-statement feb_statement.csv \
  --prior-reconciliation prior_rec.csv \
  --period-end 2025-12-31 \
  --prepared-by "J. Mensah" --prepared-date 2026-02-18 \
  --reviewed-by "" \
  --client "Fairmont Logistics" \
  --out "Fairmont - 2025 Bank Rec Review.xlsx"

Eight tests:

  • Test 1 — Independent recomputation. Bank balance ± reconciling items = adjusted bank; GL balance ± unrecorded items = adjusted book; the two must be equal. Computed from the item detail, not from the preparer's subtotals.
  • Test 2 — Bank balance agreed to the statement. An independent source.
  • Test 3 — Book balance agreed to the general ledger. The other independent source.
  • Test 4 — Every item supported. Each reconciling item traced to a document.
  • Test 5 — Subsequent clearance. Each item matched to the subsequent-period statement. Anything that did not clear is listed.
  • Test 6 — Stale items. Items appearing on the prior reconciliation with the same amount, aged, with the number of periods they have been carried.
  • Test 7 — Plug detection. Items whose description is generic, whose amount is suspiciously round, or that appear only on the reconciliation with no counterpart anywhere.
  • Test 8 — Preparation and review controls. Timeliness, whether it was reviewed, and whether the preparer also has disbursement authority.

Step 2 — Read subsequent clearance carefully

Clearance is the strongest evidence available here, but read it correctly:

  • An outstanding check that cleared in the subsequent period for the same amount: supported. Good.
  • Cleared for a different amount: the reconciliation was wrong, or the check was altered. Worth pursuing either way.
  • Did not clear at all and is more than a couple of months old: investigate. Was it ever issued? Was it delivered?
  • A deposit in transit that did not clear within days: a deposit in transit is by definition a deposit already made. One that never arrives at the bank is not timing — it is an exception, and it is the classic lapping and fictitious-receipts pattern.
  • An item that cleared before period end: it was not outstanding at all and should never have been on the reconciliation.

The script computes days-to-clear for each item and flags the ones that do not fit their claimed category.

Step 3 — Deliver

Workbook tabs:

  1. Review Summary — independent recomputation against the preparer's figures, the eight tests, the control observations, and the conclusion. The signable page.
  2. Independent Recomputation — the four-column form rebuilt from item detail, alongside what the preparer presented, with differences.
  3. Item Testing — every reconciling item: amount, category, support obtained, whether it cleared, when, days to clear, and any flag. Unsupported and uncleared items sort to the top.
  4. Uncleared Items — everything that did not clear, with age and the follow-up required.
  5. Stale Items — items carried from prior reconciliations, with the number of periods.
  6. Plug Indicators — generic descriptions, round amounts, and items with no counterpart.
  7. Control Observations — preparation timeliness, review evidence, segregation of duties.

Then, in chat: whether the recomputation agrees, whether both balances agreed to independent sources, how many items lacked support, how many did not clear, and the control observations. Lead with anything unsupported or uncleared — those are the findings; the arithmetic rarely is.

What to escalate

  • A reconciling item with no counterpart in any bank statement, before or after. This is the fabricated-item pattern and it is escalated regardless of amount.
  • A deposit in transit that never cleared.
  • An item carried on three or more consecutive reconciliations. It is not timing.
  • A reconciliation that only foots because of an item labelled generically. Report the label and the amount together — the juxtaposition is the finding.
  • An unreviewed reconciliation, or one prepared by someone with disbursement authority.
  • A reconciliation prepared long after period end. The longer the delay, the more opportunity to construct rather than reconcile.
  • Adjusting entries made to the cash account rather than to the reconciliation — these bypass the reconciliation entirely.
  • Prior-period reconciling items that disappeared without clearing. Something was removed rather than resolved, and the reconciliation still foots, which means something else absorbed it.

State each as a fact with the item reference and amount. Do not characterise intent.

Security posture

Fully local: standard library plus openpyxl. No network calls, no uploads, no telemetry.

Dependencies

pip install openpyxl

Built by Adopt AI — free to use and modify.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.